Volume 40, Number 6, December 2022
|Page(s)||1375 - 1384|
|Published online||10 February 2023|
Modulation recognition algorithm based on mixed attention prototype network
Information and Navigation College, Air Force Engineering University, Xi'an 710077, China
针对极少量带标签样本条件下的通信信号调制识别难题, 提出一种基于混合注意力原型网络的调制识别算法。综合元学习和度量学习的思想, 在原型网络框架下通过特征提取模块将信号映射至一个新的特征度量空间, 并通过比较该空间内各类原型与查询信号之间的距离确定查询信号调制样式。根据通信信号IQ分量的时序特点设计了由卷积神经网络和长短时记忆网络级联的特征提取模块, 并引入卷积注意力机制提升关键特征的权重; 采用基于Episode的训练策略, 使算法可泛化到新的信号识别任务中。仿真结果表明, 所提算法在每类信号只有5个带标签样本(5-way 5-shot)时平均识别率可达85.68%。
A modulation recognition algorithm based on mixed attention prototype network is proposed to solve the modulation recognition problem of communication signals with very few labeled samples. Combining the ideas of meta learning and metric learning, the algorithm maps the signal to a new feature metric space by feature extraction module in the prototype network framework and determines the modulation pattern of the query signal by comparing the distance between each prototype and the query signal in the space. According to the time sequence characteristics of communication signal IQ components, a feature extraction module is designed which is cascated by the convolutional neural network and long-short term memory network, and the convolutional attention mechanism is introduced to improve the weight of key features. The training strategy based on Episode is used to generalize the algorithm to new signal recognition tasks. The simulation results show that the average recognition rate of the present algorithm can reach 85.68% when there are only 5 labeled samples (5-way 5-shot) for each type of signal.
Key words: modulation recognition / prototype network / meta learning / metric learning
关键字 : 调制识别 / 原型网络 / 元学习 / 度量学习
© 2022 Journal of Northwestern Polytechnical University. All rights reserved.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.